I know how to use the AI tools for my purpose (that's why i use them), and of course, to make the impossible possible. Even if i failed to do so, it's not the decrease in productivity because without them, i don't think i can do better than the LLM.
Well, it seems you are not open for discussion. There is no reason to disparage the senior devs that participated in the study just because you don't like the results of the study. But the study happened, and it is clear: experienced developers are the ones that suffered from using LLMs.
> but it's just illogical for a junior to get benefits and the seniors don't
Experienced car drivers won't benefit from a youtube tutorial how to drive, junior car drivers might. That's similar to junior developers being potentially the ones who can benefit from the basic things that an LLM can help you with, e.g helping you with syntax and structure your thoughts and write a scaffold to get you started. Those are concerns that experienced developers don't need help with, similarly how experienced drivers don't need youtube tutorials how to shift a gear. There is nothing illogical in that premise. Do you agree?
> i don't think i can do better than the LLM
I most certainly can tell you that there are 1000s of developers that can do infinitely better than any of the current LLMs, and those developers are fairly often senior. It seems like a skill issue you mentioned in the beginning of your post might actually be on your side.
Counterpoint: the agents are the reason for the distraction.
> That has solutions they are not aware of to implement.
Counterpoint: there are no other current studies that suggest otherwise. Given the impact of LLMs on open source (net negative, maintainers are drowning in slop: e.g curl: https://daniel.haxx.se/blog/2025/07/14/death-by-a-thousand-s...) maybe it makes sense to be a bit more critical on LLM's supposed gains.
Let's see what we have so far:
- The only study to date suggests a net negative from using LLMs on experienced developers
- OSS maintainers are rejecting AI generated PRs due to low quality
- No other studies have come out so far to suggest otherwise
Based on my anecdotal experience and based on the currently available evidence, for me the conclusion is clear: LLMs and agents are mostly hype.
It's no surprise to me that devs who are accustomed to working on one thing at a time due to fast feedback loops have not learned to adapt to paralellizing their work (something that has been demonized at agile style organizations) and sit and wait on agents and start watching YouTube instead. The study reflects usage of emergent tools without training, and with regressive training on previous generation sequential processes, so I would expect these results. If there is any merit in coordinating multiple agents on slower feedback work, this study would not find it.
> i don't think i can do better than the LLM.
I could literally teach a below-average-intelligence 16-year-old how to write better code than any LLM I've ever seen - if they're interested and willing to learn.
My understanding is that a code monkey just does what they're told. All the planning and behind the scenes negotiations that the senior devs and management do is completely opaque to them.